Bayesian SpamAssassin

[David Relson]

At 02:41 PM 3/13/03, Dan Stromberg wrote:
>On Thu, 2003-03-13 at 11:25, elijah wrote:
> > On Thu, 13 Mar 2003, Barry Gould wrote:
> >
> > > SpamAssassin is now reportedly using Bayesian classification as a
> > > supplement to it's existing architecture:
> > > http://lwn.net/Articles/24297/
> >
> > one might ask, 'why bother?'. in my tests, bogofilter is hundreds of times
> > faster and catches more spam.
>
>SpamAssassin apparently seeds its database with rulebased spam, and
>allows the bayesian analysis to kick in after it's collected enough
>information to do so effectively.
>
>IOW, you apparently don't need to seed your database manually.

Hello all,

I just read the article.  Sounds like the bayesian analysis is an 
additional step in the scoring, hence adds to both processing time and 
accuracy - a plus and a minus.

'Tis good to hear that folks thing bogofilter is better :-)

Reading the comments, I was pleased to see ArticWolf's "shameless plug for 
bogofilter".  ArticWolf - who are you???

David